Linear algebra reservations, progress and results - Page 220

swellman · 2018-02-01, 00:31

- Lucas(1324) has been working on an overtasked machine, but is now cranking on 6 threads. Should be done in a few more weeks.

- C210 from 129^103+103^129 results reported today updated

- C246 from 143^58+58^143 results reported in post 2367 updated

fivemack · 2018-02-01, 09:47

There will be a fortnight or so's delay for F1361; it uses 21GB of memory which turns out to be too much on a 32GB machine on which I'm building a several-million-line C++ codebase repeatedly, so I'll have to move it to another machine which is currently busy. The job will take about a month so the delay is immaterial.

pinhodecarlos · 2018-02-02, 06:43

Quote:

Originally Posted by pinhodecarlos

Reserving 265*(9^265+1).C232, doing this for Paul.

LA underway, ETA 150 hours.

RichD · 2018-02-02, 17:08

Looks like the GCW number (9_266+.C255) in the 14e queue can use a few more relations. Currently the 32-bit job is at 345M.

pinhodecarlos · 2018-02-02, 20:21

Quote:

Originally Posted by RichD

Looks like the GCW number (9_266+.C255) in the 14e queue can use a few more relations. Currently the 32-bit job is at 345M.

That’s why I’ve chosen to post process the one queued on the 15e queue. Paul needs to estimate more relations up to ~450M.

VBCurtis · 2018-02-02, 22:57

F1469 is complete:

Code:

prp77 factor: 14707318795142178793712842139214249585232198901328179403254741435264239387637
prp107 factor: 12238715020863597403126410370715840499161703793870153678980148848432866729077776897334456680693270487543061

Log at https://pastebin.com/H2afkbvG

After remdups, roughly 385M unique relations remained. TD 134 produced a 20.1M matrix of density 71. So, I reduced relations available:
375M relations produced 19.35M of density 87
360M relations produced 18.8M of density 99
345M relations produced 18.5M of density 109
330M relations produced 17.3M of density 134
332M relations produced 17.22M of density 134
I used the last matrix. It consumed 9.1GB memory according to "top".

I don't think any 32-bit job needs more than 450M raw relations, and many of these will build nice matrices with 425M. We're oversieving by 15% beyond the relation set that will build 130+ density matrices, and inattentive postprocessors will waste substantial time solving the bigger matrices that result.

swellman · 2018-02-02, 23:58

Quote:

Originally Posted by VBCurtis

F1469 is complete:

Code:

prp77 factor: 14707318795142178793712842139214249585232198901328179403254741435264239387637
prp107 factor: 12238715020863597403126410370715840499161703793870153678980148848432866729077776897334456680693270487543061

Log at https://pastebin.com/H2afkbvG

After remdups, roughly 385M unique relations remained. TD 134 produced a 20.1M matrix of density 71. So, I reduced relations available:
375M relations produced 19.35M of density 87
360M relations produced 18.8M of density 99
345M relations produced 18.5M of density 109
330M relations produced 17.3M of density 134
332M relations produced 17.22M of density 134
I used the last matrix. It consumed 9.1GB memory according to "top".

I don't think any 32-bit job needs more than 450M raw relations, and many of these will build nice matrices with 425M. We're oversieving by 15% beyond the relation set that will build 130+ density matrices, and inattentive postprocessors will waste substantial time solving the bigger matrices that result.

Nice finding, consistent with your earlier work. I know that when posting “my” jobs the target number of relations is always high, 490-500M for 32-bit jobs. Future requests from me will target 450M going forward (I have few 31-bit jobs in the hopper, all GNFS C175-182 currently).

VBCurtis · 2018-02-03, 02:20

Quote:

Originally Posted by swellman

Nice finding, consistent with your earlier work. I know that when posting “my” jobs the target number of relations is always high, 490-500M for 32-bit jobs. Future requests from me will target 450M going forward (I have few 31-bit jobs in the hopper, all GNFS C175-182 currently).

It's worth pointing out for posterity that this job began with 484M raw relations; my log does not show that since remdups was run before filtering. The duplication rate of the highest/last Q is usually higher than the lower Q values, so it's reasonable to think that 410M raw relations would have yielded the 330-335M unique relations needed for the desired density of 134. This was a GNFS-183 job, so 175-180 jobs should aim for 400M or 410M raw relations.
This poly may have had unusually low duplication of relations, but it's a safe bet that a smaller job won't need more raw relations.

swellman · 2018-02-03, 16:18

Reserving 144^55+55^144 C180 cofactor for postprocessing. Thank you.

swellman · 2018-02-04, 00:50

C231_134_79 factored

Code:

prp63 factor: 121228617008936764925388482653465206046556191972729649438307283
prp169 factor: 5583421281472380925687498948836078361630770800617356324851941804728158677443434969099749217012072569073397018190430180207527906857723918495624214642334514406145094063491

570995583 raw / 436934597 unique relations

fivemack · 2018-02-04, 17:56

Taking L1384 (ETA 20 February)

2018-02-01, 00:31	#2410
swellman Jun 2012 11×281 Posts	Update - Lucas(1324) has been working on an overtasked machine, but is now cranking on 6 threads. Should be done in a few more weeks. - C210 from 129^103+103^129 results reported today updated - C246 from 143^58+58^143 results reported in post 2367 updated Last fiddled with by fivemack on 2018-02-01 at 09:50

2018-02-02, 22:57	#2415
VBCurtis "Curtis" Feb 2005 Riverside, CA 11×443 Posts	F1469 is complete: Code: prp77 factor: 14707318795142178793712842139214249585232198901328179403254741435264239387637 prp107 factor: 12238715020863597403126410370715840499161703793870153678980148848432866729077776897334456680693270487543061 Log at https://pastebin.com/H2afkbvG After remdups, roughly 385M unique relations remained. TD 134 produced a 20.1M matrix of density 71. So, I reduced relations available: 375M relations produced 19.35M of density 87 360M relations produced 18.8M of density 99 345M relations produced 18.5M of density 109 330M relations produced 17.3M of density 134 332M relations produced 17.22M of density 134 I used the last matrix. It consumed 9.1GB memory according to "top". I don't think any 32-bit job needs more than 450M raw relations, and many of these will build nice matrices with 425M. We're oversieving by 15% beyond the relation set that will build 130+ density matrices, and inattentive postprocessors will waste substantial time solving the bigger matrices that result.

2018-02-04, 17:56	#2420
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 2³×11×73 Posts	Taking L1384 (ETA 20 February) Last fiddled with by fivemack on 2018-02-07 at 21:56

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
restarting nfs linear algebra	cubaq	YAFU	2	2017-04-02 11:35
Linear algebra at 600%	CRGreathouse	Msieve	8	2009-08-05 07:25
Linear algebra crashes	10metreh	Msieve	3	2009-02-02 08:34
Linear algebra proof	Damian	Math	8	2007-02-12 22:25
Linear algebra in MPQS	R1zZ1	Factoring	2	2007-02-02 06:45

2018-02-01, 09:47	#2411
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 6424₁₀ Posts	There will be a fortnight or so's delay for F1361; it uses 21GB of memory which turns out to be too much on a 32GB machine on which I'm building a several-million-line C++ codebase repeatedly, so I'll have to move it to another machine which is currently busy. The job will take about a month so the delay is immaterial.

2018-02-02, 17:08	#2413
RichD Sep 2008 Kansas 2·3·5·113 Posts	Looks like the GCW number (9_266+.C255) in the 14e queue can use a few more relations. Currently the 32-bit job is at 345M.

2018-02-03, 16:18	#2418
swellman Jun 2012 110000010011₂ Posts	Reserving 144^55+55^144 C180 cofactor for postprocessing. Thank you.